djc_poss

数据挖掘中的一算法 ines算法 c下实现的。适合初学习数据挖掘者借鉴

#!/bin/sh
#-----------------------------------------------------------------------
# File    : djc_poss
# Contents: possibilistic network induction on Danish Jersey Cattle data
#           on real world data
# Author  : Christian Borgelt
# History : 17.12.1999 file created
#           16.01.2000 simulated annealing added
#           09.03.2002 shell changed from csh to sh
#           10.04.2002 all induction loops moved into functions
#-----------------------------------------------------------------------

function collect () {
gawk '
/evaluation of/         { network = $3  }
/number of attributes/  { attcnt  = $NF }
/number of conditions/  { concnt  = $NF }
/number of parameters/  { parcnt  = $NF }
/number of tuples/      { tplcnt  = $NF }
/impossible tuples/     { imptpl  = $4  }
/minimum/               { minimum = $NF }
/average/               { average = $NF }
/maximum/               { maximum = $NF }
/additional conditions/ { addcnt  = $NF }
/missing    conditions/ { miscnt  = $NF }
END {
  printf("%-12s", network);
  printf(" %3d %5d", concnt, parcnt);
  printf(" %10g %10g %10g", average, minimum, maximum);
}'
}

#-----------------------------------------------------------------------

function average () {
gawk '
function output() {
  if (NR > 0) {
    printf("%-10s", network);
    printf(" %6.1f %7.1f", concnt/n, parcnt/n);
    printf(" %7.3f/%6.3f/%6.3f",   avg1/n, min1/n, max1/n);
    printf(" %7.3f/%6.3f/%6.3f\n", avg2/n, min2/n, max2/n);
  }
}
BEGIN { network = ""; n = 0; }
($1 == network) {
  concnt += $2; parcnt += $3;
  avg1   += $4; min1   += $5; max1 += $6;
  avg2   += $7; min2   += $8; max2 += $9;
  n++;
}
($1 != network) {
  if (n > 0) output();
  network = $1; n = 1;
  concnt  = $2; parcnt = $3;
  avg1    = $4; min1   = $5; max1 = $6;
  avg2    = $7; min2   = $8; max2 = $9;
}
END { if (n > 0) output(); }' poss.tmp
}

#-----------------------------------------------------------------------

function evaluate ()
{                               # --- evaluate a given network
  neval -u"*" -c djc.pnt $1 djc.tab 2> /dev/null | \
    collect >> poss.tmp
  neval -u"*" -s         $1 djc.tab 2> /dev/null | \
    gawk '/average/ { printf(" %10g",   $NF); }
          /minimum/ { printf(" %10g",   $NF); }
          /maximum/ { printf(" %10g\n", $NF); }' >> poss.tmp
  rm -f $1
}  # evaluate()

#-----------------------------------------------------------------------

function induce ()
{                               # --- induce and evaluate networks
  rm -f poss.tmp
  ines -u'*' -mx -s$1 -e$2 $3 djc.dom djc.tab $2 2> /dev/null
  evaluate $2
  average | tee -a poss.res
}  # induce()

#-----------------------------------------------------------------------

function indlocal ()
{                               # --- induce and evaluate networks
  rm -f poss.tmp
  ines -u'*' -mx -s$1 -e$2 djc.dom djc.tab $2 2> /dev/null
  ines -c0 -g-1e-12 -u'*' -mx -s$1 -e$2 $2 djc.tab $2 2> /dev/null
  evaluate $2
  average | tee -a poss.res
}  # indlocal()

#-----------------------------------------------------------------------

function fixed ()
{                               # --- evaluate empty/original network
  rm -f poss.tmp
  if [[ $1 == indep ]]; then in="djc.dom";
                        else in="djc.pnt"; fi
  ines  -u'*' -mx $in djc.tab $1 2> /dev/null
  evaluate $1
  average | tee -a poss.res
}  # fixed()

#-----------------------------------------------------------------------

function owst ()
{                               # --- optimum weight spanning tree cons.
  echo "---owst--------------------------------------------------------------"\
       | tee -a poss.res
  for m in spcgain spcsgr1 chi2 mutspc; do
    induce owst $m
  done
}  # owst()

#-----------------------------------------------------------------------

function extst ()
{                               # --- optimum weight spanning tree ext.
  echo "---extst-------------------------------------------------------------"\
       | tee -a poss.res
  for m in spcgain spcsgr1 chi2 mutspc; do
    induce extst $m
  done
}  # owst()

#-----------------------------------------------------------------------

function topord ()
{                               # --- selection on topological order
  echo "---topord------------------------------------------------------------"\
       | tee -a poss.res
  for m in spcgain spcgr spcsgr1 chi2 mutspc; do
    induce topord $m
  done
}  # topord()

#-----------------------------------------------------------------------

function noloop ()
{                               # --- selection avoiding directed loops
  echo "---noloop------------------------------------------------------------"\
       | tee -a poss.res
  for m in spcgain spcsgr1 chi2 mutspc; do
    ines  -u'*' -mx -snoloop -e$m djc.dom djc.tab $m 2> /dev/null
    neval -u'*' $m djc.tab 2> /dev/null | collect > poss.tmp
    rm -f $m
    average | tee -a poss.res
  done
}  # noloop()

#-----------------------------------------------------------------------

function local ()
{                               # --- local structure learning
  echo "---local-------------------------------------------------------------"\
       | tee -a poss.res
  for m in spcgain spcgr spcsgr1 chi2 mutspc; do
    induce topord $m -g-1e-12
  done
}  # local()

#-----------------------------------------------------------------------

function lwise ()
{                               # --- local structure learning
  echo "---lwise-------------------------------------------------------------"\
       | tee -a poss.res
  for m in spcgain spcgr spcsgr1 chi2 mutspc; do
    induce topord $m -G-1e-12
  done
}  # lwise()

#-----------------------------------------------------------------------

function post ()
{                               # --- local structure learning
  echo "---post--------------------------------------------------------------"\
       | tee -a poss.res
  for m in spcgain spcgr spcsgr1 chi2 mutspc; do
    indlocal topord $m
  done
}  # post()

#-----------------------------------------------------------------------

function sian ()
{                               # --- hypertree simulated annealing
  echo "---sian--------------------------------------------------------------"\
       | tee -a poss.res
  for p in 0 001; do   # 0005
    if (( p == 0 )); then out="sian_no";
                     else out="sian_yes"; fi
    rm -f poss.tmp
    for (( i = 0; i < 10; i++ )); do
      ines  -u'*' -mx -ssian -w0.$p -S1$i djc.dom djc.tab $out \
        2> /dev/null
      evaluate $out
    done
    average | tee -a poss.res
  done
}  # sian()

#-----------------------------------------------------------------------

echo "                                 absolute              relative"\
     | tee poss.res
echo "network      cond  params    avg    min    max     avg    min    max"\
     | tee -a poss.res
echo "---------------------------------------------------------------------"\
     | tee -a poss.res

fixed indep             # evaluate empty    network
fixed orig              # evaluate original network

owst                    # optimum weight spanning tree construction
#extst                   # optimum weight spanning tree extension
topord                  # condition selection on topological order
#noloop                  # condition selection avoiding directed loops
local                   # local structure learning (unrestricted)
#lwise                   # local structure learning (levelwise)
#post                    # local structure learning (preserving global)
sian                    # hypertree simulated annealing

rm -f poss.tmp          # clean up temporary file